src/gromacs/gmxlib/nonbonded/nb_kernel_sse2_double/nb_kernel_ElecCoul_VdwLJ_GeomP1P1_sse2_double.c

   1 /*
   2  * Note: this file was generated by the Gromacs sse2_double kernel generator.
   3  *
   4  *                This source code is part of
   5  *
   6  *                 G   R   O   M   A   C   S
   7  *
   8  * Copyright (c) 2001-2012, The GROMACS Development Team
   9  *
  10  * Gromacs is a library for molecular simulation and trajectory analysis,
  11  * written by Erik Lindahl, David van der Spoel, Berk Hess, and others - for
  12  * a full list of developers and information, check out http://www.gromacs.org
  13  *
  14  * This program is free software; you can redistribute it and/or modify it under
  15  * the terms of the GNU Lesser General Public License as published by the Free
  16  * Software Foundation; either version 2 of the License, or (at your option) any
  17  * later version.
  18  *
  19  * To help fund GROMACS development, we humbly ask that you cite
  20  * the papers people have written on it - you can find them on the website.
  21  */
  22 #ifdef HAVE_CONFIG_H
  23 #include <config.h>
  24 #endif
  25
  26 #include <math.h>
  27
  28 #include "../nb_kernel.h"
  29 #include "types/simple.h"
  30 #include "vec.h"
  31 #include "nrnb.h"
  32
  33 #include "gmx_math_x86_sse2_double.h"
  34 #include "kernelutil_x86_sse2_double.h"
  35
  36 /*
  37  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwLJ_GeomP1P1_VF_sse2_double
  38  * Electrostatics interaction: Coulomb
  39  * VdW interaction:            LennardJones
  40  * Geometry:                   Particle-Particle
  41  * Calculate force/pot:        PotentialAndForce
  42  */
  43 void
  44 nb_kernel_ElecCoul_VdwLJ_GeomP1P1_VF_sse2_double
  45                     (t_nblist * gmx_restrict                nlist,
  46                      rvec * gmx_restrict                    xx,
  47                      rvec * gmx_restrict                    ff,
  48                      t_forcerec * gmx_restrict              fr,
  49                      t_mdatoms * gmx_restrict               mdatoms,
  50                      nb_kernel_data_t * gmx_restrict        kernel_data,
  51                      t_nrnb * gmx_restrict                  nrnb)
  52 {
  53     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
  54      * just 0 for non-waters.
  55      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
  56      * jnr indices corresponding to data put in the four positions in the SIMD register.
  57      */
  58     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
  59     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
  60     int              jnrA,jnrB;
  61     int              j_coord_offsetA,j_coord_offsetB;
  62     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
  63     real             rcutoff_scalar;
  64     real             *shiftvec,*fshift,*x,*f;
  65     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
  66     int              vdwioffset0;
  67     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
  68     int              vdwjidx0A,vdwjidx0B;
  69     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
  70     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
  71     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
  72     real             *charge;
  73     int              nvdwtype;
  74     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
  75     int              *vdwtype;
  76     real             *vdwparam;
  77     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
  78     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
  79     __m128d          dummy_mask,cutoff_mask;
  80     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
  81     __m128d          one     = _mm_set1_pd(1.0);
  82     __m128d          two     = _mm_set1_pd(2.0);
  83     x                = xx[0];
  84     f                = ff[0];
  85
  86     nri              = nlist->nri;
  87     iinr             = nlist->iinr;
  88     jindex           = nlist->jindex;
  89     jjnr             = nlist->jjnr;
  90     shiftidx         = nlist->shift;
  91     gid              = nlist->gid;
  92     shiftvec         = fr->shift_vec[0];
  93     fshift           = fr->fshift[0];
  94     facel            = _mm_set1_pd(fr->epsfac);
  95     charge           = mdatoms->chargeA;
  96     nvdwtype         = fr->ntype;
  97     vdwparam         = fr->nbfp;
  98     vdwtype          = mdatoms->typeA;
  99
 100     /* Avoid stupid compiler warnings */
 101     jnrA = jnrB = 0;
 102     j_coord_offsetA = 0;
 103     j_coord_offsetB = 0;
 104
 105     outeriter        = 0;
 106     inneriter        = 0;
 107
 108     /* Start outer loop over neighborlists */
 109     for(iidx=0; iidx<nri; iidx++)
 110     {
 111         /* Load shift vector for this list */
 112         i_shift_offset   = DIM*shiftidx[iidx];
 113
 114         /* Load limits for loop over neighbors */
 115         j_index_start    = jindex[iidx];
 116         j_index_end      = jindex[iidx+1];
 117
 118         /* Get outer coordinate index */
 119         inr              = iinr[iidx];
 120         i_coord_offset   = DIM*inr;
 121
 122         /* Load i particle coords and add shift vector */
 123         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 124
 125         fix0             = _mm_setzero_pd();
 126         fiy0             = _mm_setzero_pd();
 127         fiz0             = _mm_setzero_pd();
 128
 129         /* Load parameters for i particles */
 130         iq0              = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
 131         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 132
 133         /* Reset potential sums */
 134         velecsum         = _mm_setzero_pd();
 135         vvdwsum          = _mm_setzero_pd();
 136
 137         /* Start inner kernel loop */
 138         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 139         {
 140
 141             /* Get j neighbor index, and coordinate index */
 142             jnrA             = jjnr[jidx];
 143             jnrB             = jjnr[jidx+1];
 144             j_coord_offsetA  = DIM*jnrA;
 145             j_coord_offsetB  = DIM*jnrB;
 146
 147             /* load j atom coordinates */
 148             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 149                                               &jx0,&jy0,&jz0);
 150
 151             /* Calculate displacement vector */
 152             dx00             = _mm_sub_pd(ix0,jx0);
 153             dy00             = _mm_sub_pd(iy0,jy0);
 154             dz00             = _mm_sub_pd(iz0,jz0);
 155
 156             /* Calculate squared distance and things based on it */
 157             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 158
 159             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 160
 161             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 162
 163             /* Load parameters for j particles */
 164             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 165             vdwjidx0A        = 2*vdwtype[jnrA+0];
 166             vdwjidx0B        = 2*vdwtype[jnrB+0];
 167
 168             /**************************
 169              * CALCULATE INTERACTIONS *
 170              **************************/
 171
 172             /* Compute parameters for interactions between i and j atoms */
 173             qq00             = _mm_mul_pd(iq0,jq0);
 174             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 175                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 176
 177             /* COULOMB ELECTROSTATICS */
 178             velec            = _mm_mul_pd(qq00,rinv00);
 179             felec            = _mm_mul_pd(velec,rinvsq00);
 180
 181             /* LENNARD-JONES DISPERSION/REPULSION */
 182
 183             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 184             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 185             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 186             vvdw             = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
 187             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 188
 189             /* Update potential sum for this i atom from the interaction with this j atom. */
 190             velecsum         = _mm_add_pd(velecsum,velec);
 191             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 192
 193             fscal            = _mm_add_pd(felec,fvdw);
 194
 195             /* Calculate temporary vectorial force */
 196             tx               = _mm_mul_pd(fscal,dx00);
 197             ty               = _mm_mul_pd(fscal,dy00);
 198             tz               = _mm_mul_pd(fscal,dz00);
 199
 200             /* Update vectorial force */
 201             fix0             = _mm_add_pd(fix0,tx);
 202             fiy0             = _mm_add_pd(fiy0,ty);
 203             fiz0             = _mm_add_pd(fiz0,tz);
 204
 205             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
 206
 207             /* Inner loop uses 40 flops */
 208         }
 209
 210         if(jidx<j_index_end)
 211         {
 212
 213             jnrA             = jjnr[jidx];
 214             j_coord_offsetA  = DIM*jnrA;
 215
 216             /* load j atom coordinates */
 217             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 218                                               &jx0,&jy0,&jz0);
 219
 220             /* Calculate displacement vector */
 221             dx00             = _mm_sub_pd(ix0,jx0);
 222             dy00             = _mm_sub_pd(iy0,jy0);
 223             dz00             = _mm_sub_pd(iz0,jz0);
 224
 225             /* Calculate squared distance and things based on it */
 226             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 227
 228             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 229
 230             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 231
 232             /* Load parameters for j particles */
 233             jq0              = _mm_load_sd(charge+jnrA+0);
 234             vdwjidx0A        = 2*vdwtype[jnrA+0];
 235
 236             /**************************
 237              * CALCULATE INTERACTIONS *
 238              **************************/
 239
 240             /* Compute parameters for interactions between i and j atoms */
 241             qq00             = _mm_mul_pd(iq0,jq0);
 242             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 243
 244             /* COULOMB ELECTROSTATICS */
 245             velec            = _mm_mul_pd(qq00,rinv00);
 246             felec            = _mm_mul_pd(velec,rinvsq00);
 247
 248             /* LENNARD-JONES DISPERSION/REPULSION */
 249
 250             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 251             vvdw6            = _mm_mul_pd(c6_00,rinvsix);
 252             vvdw12           = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
 253             vvdw             = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
 254             fvdw             = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
 255
 256             /* Update potential sum for this i atom from the interaction with this j atom. */
 257             velec            = _mm_unpacklo_pd(velec,_mm_setzero_pd());
 258             velecsum         = _mm_add_pd(velecsum,velec);
 259             vvdw             = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
 260             vvdwsum          = _mm_add_pd(vvdwsum,vvdw);
 261
 262             fscal            = _mm_add_pd(felec,fvdw);
 263
 264             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 265
 266             /* Calculate temporary vectorial force */
 267             tx               = _mm_mul_pd(fscal,dx00);
 268             ty               = _mm_mul_pd(fscal,dy00);
 269             tz               = _mm_mul_pd(fscal,dz00);
 270
 271             /* Update vectorial force */
 272             fix0             = _mm_add_pd(fix0,tx);
 273             fiy0             = _mm_add_pd(fiy0,ty);
 274             fiz0             = _mm_add_pd(fiz0,tz);
 275
 276             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
 277
 278             /* Inner loop uses 40 flops */
 279         }
 280
 281         /* End of innermost loop */
 282
 283         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 284                                               f+i_coord_offset,fshift+i_shift_offset);
 285
 286         ggid                        = gid[iidx];
 287         /* Update potential energies */
 288         gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
 289         gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
 290
 291         /* Increment number of inner iterations */
 292         inneriter                  += j_index_end - j_index_start;
 293
 294         /* Outer loop uses 9 flops */
 295     }
 296
 297     /* Increment number of outer iterations */
 298     outeriter        += nri;
 299
 300     /* Update outer/inner flops */
 301
 302     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_VF,outeriter*9 + inneriter*40);
 303 }
 304 /*
 305  * Gromacs nonbonded kernel:   nb_kernel_ElecCoul_VdwLJ_GeomP1P1_F_sse2_double
 306  * Electrostatics interaction: Coulomb
 307  * VdW interaction:            LennardJones
 308  * Geometry:                   Particle-Particle
 309  * Calculate force/pot:        Force
 310  */
 311 void
 312 nb_kernel_ElecCoul_VdwLJ_GeomP1P1_F_sse2_double
 313                     (t_nblist * gmx_restrict                nlist,
 314                      rvec * gmx_restrict                    xx,
 315                      rvec * gmx_restrict                    ff,
 316                      t_forcerec * gmx_restrict              fr,
 317                      t_mdatoms * gmx_restrict               mdatoms,
 318                      nb_kernel_data_t * gmx_restrict        kernel_data,
 319                      t_nrnb * gmx_restrict                  nrnb)
 320 {
 321     /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
 322      * just 0 for non-waters.
 323      * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
 324      * jnr indices corresponding to data put in the four positions in the SIMD register.
 325      */
 326     int              i_shift_offset,i_coord_offset,outeriter,inneriter;
 327     int              j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
 328     int              jnrA,jnrB;
 329     int              j_coord_offsetA,j_coord_offsetB;
 330     int              *iinr,*jindex,*jjnr,*shiftidx,*gid;
 331     real             rcutoff_scalar;
 332     real             *shiftvec,*fshift,*x,*f;
 333     __m128d          tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
 334     int              vdwioffset0;
 335     __m128d          ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
 336     int              vdwjidx0A,vdwjidx0B;
 337     __m128d          jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
 338     __m128d          dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
 339     __m128d          velec,felec,velecsum,facel,crf,krf,krf2;
 340     real             *charge;
 341     int              nvdwtype;
 342     __m128d          rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
 343     int              *vdwtype;
 344     real             *vdwparam;
 345     __m128d          one_sixth   = _mm_set1_pd(1.0/6.0);
 346     __m128d          one_twelfth = _mm_set1_pd(1.0/12.0);
 347     __m128d          dummy_mask,cutoff_mask;
 348     __m128d          signbit   = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
 349     __m128d          one     = _mm_set1_pd(1.0);
 350     __m128d          two     = _mm_set1_pd(2.0);
 351     x                = xx[0];
 352     f                = ff[0];
 353
 354     nri              = nlist->nri;
 355     iinr             = nlist->iinr;
 356     jindex           = nlist->jindex;
 357     jjnr             = nlist->jjnr;
 358     shiftidx         = nlist->shift;
 359     gid              = nlist->gid;
 360     shiftvec         = fr->shift_vec[0];
 361     fshift           = fr->fshift[0];
 362     facel            = _mm_set1_pd(fr->epsfac);
 363     charge           = mdatoms->chargeA;
 364     nvdwtype         = fr->ntype;
 365     vdwparam         = fr->nbfp;
 366     vdwtype          = mdatoms->typeA;
 367
 368     /* Avoid stupid compiler warnings */
 369     jnrA = jnrB = 0;
 370     j_coord_offsetA = 0;
 371     j_coord_offsetB = 0;
 372
 373     outeriter        = 0;
 374     inneriter        = 0;
 375
 376     /* Start outer loop over neighborlists */
 377     for(iidx=0; iidx<nri; iidx++)
 378     {
 379         /* Load shift vector for this list */
 380         i_shift_offset   = DIM*shiftidx[iidx];
 381
 382         /* Load limits for loop over neighbors */
 383         j_index_start    = jindex[iidx];
 384         j_index_end      = jindex[iidx+1];
 385
 386         /* Get outer coordinate index */
 387         inr              = iinr[iidx];
 388         i_coord_offset   = DIM*inr;
 389
 390         /* Load i particle coords and add shift vector */
 391         gmx_mm_load_shift_and_1rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,&ix0,&iy0,&iz0);
 392
 393         fix0             = _mm_setzero_pd();
 394         fiy0             = _mm_setzero_pd();
 395         fiz0             = _mm_setzero_pd();
 396
 397         /* Load parameters for i particles */
 398         iq0              = _mm_mul_pd(facel,_mm_load1_pd(charge+inr+0));
 399         vdwioffset0      = 2*nvdwtype*vdwtype[inr+0];
 400
 401         /* Start inner kernel loop */
 402         for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
 403         {
 404
 405             /* Get j neighbor index, and coordinate index */
 406             jnrA             = jjnr[jidx];
 407             jnrB             = jjnr[jidx+1];
 408             j_coord_offsetA  = DIM*jnrA;
 409             j_coord_offsetB  = DIM*jnrB;
 410
 411             /* load j atom coordinates */
 412             gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
 413                                               &jx0,&jy0,&jz0);
 414
 415             /* Calculate displacement vector */
 416             dx00             = _mm_sub_pd(ix0,jx0);
 417             dy00             = _mm_sub_pd(iy0,jy0);
 418             dz00             = _mm_sub_pd(iz0,jz0);
 419
 420             /* Calculate squared distance and things based on it */
 421             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 422
 423             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 424
 425             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 426
 427             /* Load parameters for j particles */
 428             jq0              = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
 429             vdwjidx0A        = 2*vdwtype[jnrA+0];
 430             vdwjidx0B        = 2*vdwtype[jnrB+0];
 431
 432             /**************************
 433              * CALCULATE INTERACTIONS *
 434              **************************/
 435
 436             /* Compute parameters for interactions between i and j atoms */
 437             qq00             = _mm_mul_pd(iq0,jq0);
 438             gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
 439                                          vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
 440
 441             /* COULOMB ELECTROSTATICS */
 442             velec            = _mm_mul_pd(qq00,rinv00);
 443             felec            = _mm_mul_pd(velec,rinvsq00);
 444
 445             /* LENNARD-JONES DISPERSION/REPULSION */
 446
 447             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 448             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
 449
 450             fscal            = _mm_add_pd(felec,fvdw);
 451
 452             /* Calculate temporary vectorial force */
 453             tx               = _mm_mul_pd(fscal,dx00);
 454             ty               = _mm_mul_pd(fscal,dy00);
 455             tz               = _mm_mul_pd(fscal,dz00);
 456
 457             /* Update vectorial force */
 458             fix0             = _mm_add_pd(fix0,tx);
 459             fiy0             = _mm_add_pd(fiy0,ty);
 460             fiz0             = _mm_add_pd(fiz0,tz);
 461
 462             gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,tx,ty,tz);
 463
 464             /* Inner loop uses 34 flops */
 465         }
 466
 467         if(jidx<j_index_end)
 468         {
 469
 470             jnrA             = jjnr[jidx];
 471             j_coord_offsetA  = DIM*jnrA;
 472
 473             /* load j atom coordinates */
 474             gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
 475                                               &jx0,&jy0,&jz0);
 476
 477             /* Calculate displacement vector */
 478             dx00             = _mm_sub_pd(ix0,jx0);
 479             dy00             = _mm_sub_pd(iy0,jy0);
 480             dz00             = _mm_sub_pd(iz0,jz0);
 481
 482             /* Calculate squared distance and things based on it */
 483             rsq00            = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
 484
 485             rinv00           = gmx_mm_invsqrt_pd(rsq00);
 486
 487             rinvsq00         = _mm_mul_pd(rinv00,rinv00);
 488
 489             /* Load parameters for j particles */
 490             jq0              = _mm_load_sd(charge+jnrA+0);
 491             vdwjidx0A        = 2*vdwtype[jnrA+0];
 492
 493             /**************************
 494              * CALCULATE INTERACTIONS *
 495              **************************/
 496
 497             /* Compute parameters for interactions between i and j atoms */
 498             qq00             = _mm_mul_pd(iq0,jq0);
 499             gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
 500
 501             /* COULOMB ELECTROSTATICS */
 502             velec            = _mm_mul_pd(qq00,rinv00);
 503             felec            = _mm_mul_pd(velec,rinvsq00);
 504
 505             /* LENNARD-JONES DISPERSION/REPULSION */
 506
 507             rinvsix          = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
 508             fvdw             = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
 509
 510             fscal            = _mm_add_pd(felec,fvdw);
 511
 512             fscal            = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
 513
 514             /* Calculate temporary vectorial force */
 515             tx               = _mm_mul_pd(fscal,dx00);
 516             ty               = _mm_mul_pd(fscal,dy00);
 517             tz               = _mm_mul_pd(fscal,dz00);
 518
 519             /* Update vectorial force */
 520             fix0             = _mm_add_pd(fix0,tx);
 521             fiy0             = _mm_add_pd(fiy0,ty);
 522             fiz0             = _mm_add_pd(fiz0,tz);
 523
 524             gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,tx,ty,tz);
 525
 526             /* Inner loop uses 34 flops */
 527         }
 528
 529         /* End of innermost loop */
 530
 531         gmx_mm_update_iforce_1atom_swizzle_pd(fix0,fiy0,fiz0,
 532                                               f+i_coord_offset,fshift+i_shift_offset);
 533
 534         /* Increment number of inner iterations */
 535         inneriter                  += j_index_end - j_index_start;
 536
 537         /* Outer loop uses 7 flops */
 538     }
 539
 540     /* Increment number of outer iterations */
 541     outeriter        += nri;
 542
 543     /* Update outer/inner flops */
 544
 545     inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_F,outeriter*7 + inneriter*34);
 546 }